Many of their phone-specific pages cite the manufacturer as the only data source. This includes a phone I'm playing with at the moment, which happens to have one of the worst SAR ratings on the ewg.org list. (Worse than the Blackberry.) I followed their link, and it brought me to a user manual, which did in fact show the same values shown on the list.
Call me paranoid, but that didn't really satisfy me. For one thing, I don't trust user manuals all that much when it comes to fine details that might have changed since they were written. For another, this phone supports several different radio frequencies, including Wi-Fi and several different GSM bands, yet the manual and ewg.org fail to reflect this with multiple SAR values. So, I looked up the FCC ID for my phone and followed it to the FCC's radiation report on that model. What I found was much more informative.
As you might expect, the FCC's SAR measurements showed quite a range of values, depending on which radio is in use, which channel is in use, and how the phone is held. According to this data, my particular phone habits and service provider should yield around half the SAR that was reported by ewg.org, comparable to their best-rated models.
This exercise was interesting, and set my mind at ease a little, but I'm still going to use a wired headset whenever possible. Again, call me paranoid if you like. There simply hasn't been enough time for us to observe the long-term effects of having a microwave broadcast antenna plastered to our heads, and I don't trust studies that claim all is well when they're funded by the cell phone industry.
Some of you might find this US Senate hearing interesting:
http://appropriations.senate.gov/webcasts.cfm?method=webcasts.view&id=2a7f2e87-68a0-48a3-b16b-08ac1b98cc42
http://www.c-spanarchives.org/program/288879-1
http://www.mapcruzin.com/news/cell-phone-health-effects-hearing.htm